CODEOWNERS

-Original file line number
+Diff line change
@@ Expand Up / @@ -47,6 +47,7 @@ @@
     /ggml/cmake/                            @ggerganov
     /ggml/include/                          @ggerganov
     /ggml/src/ggml-common.h                 @ggerganov
+    /ggml/src/ggml-blas/                    @taronaeo
     /ggml/src/ggml-cpu/                     @ggerganov
     /ggml/src/ggml-cpu/spacemit/            @alex-spacemit
     /ggml/src/ggml-cuda/fattn*              @JohannesGaessler
@@ Expand Down @@

ggml/src/ggml-blas/CMakeLists.txt

-Original file line number
+Diff line change
@@ Expand Up / @@ -11,9 +11,10 @@ find_package(BLAS) @@
     if (BLAS_FOUND)
         message(STATUS "BLAS found, Libraries: ${BLAS_LIBRARIES}")
-        ggml_add_backend_library(ggml-blas
-                                 ggml-blas.cpp
-                                )
+        file(GLOB GGML_SOURCES_BLAS "*.c" "*.cpp")
+        file(GLOB GGML_HEADERS_BLAS "*.h" "*.hpp")
+        ggml_add_backend_library(ggml-blas ${GGML_HEADERS_BLAS} ${GGML_SOURCES_BLAS})
         if (${GGML_BLAS_VENDOR} MATCHES "Apple")
             add_compile_definitions(ACCELERATE_NEW_LAPACK)
@@ Expand Down @@

ggml/src/ggml-blas/common.hpp

-Original file line number
+Diff line change
@@ -0,0 +1,67 @@
+    #pragma once
+    #include "ggml.h"
+    #include "ggml-impl.h"
+    #include "ggml-backend-impl.h"
+    #include <vector>
+    #include <memory>
+    #include <future>
+    #if defined(GGML_BLAS_USE_ACCELERATE)
+    #   include <Accelerate/Accelerate.h>
+    #elif defined(GGML_BLAS_USE_MKL)
+    #   include <mkl.h>
+    #elif defined(GGML_BLAS_USE_BLIS)
+    #   include <blis.h>
+    #elif defined(GGML_BLAS_USE_NVPL)
+    #   include <nvpl_blas.h>
+    #else
+    #   include <cblas.h>
+    #endif
+    #define GGML_BLAS_NAME    "BLAS"
+    #define GGML_BLAS_VERSION GGML_BACKEND_API_VERSION
+    #ifdef __cplusplus
+    extern "C" {
+    #endif
+    struct ggml_backend_blas_buffer {
+        void * data;  // dequantized data
+        size_t size;  // ggml_nelements * sizeof(float)
+    };
+    struct ggml_backend_blas_buffer_context {
+        void * data;
+        size_t size;
+        std::vector<ggml_backend_blas_buffer *> buffers;
+        ~ggml_backend_blas_buffer_context() {
+            ggml_aligned_free(data, size);
+            for (auto * extra : buffers) {
+                ggml_aligned_free(extra->data, extra->size);
+                delete extra;
+            }
+        }
+    };
+    struct ggml_backend_blas_buffer_type_context {
+        int n_threads;
+    #ifndef GGML_USE_OPENMP
+        std::vector<std::future<void>> tasks;
+    #endif
+    };
+    struct ggml_backend_blas_context {
+        int n_threads;
+    };
+    struct ggml_backend_blas_device_context {
+        char _dummy;  // Prevent empty struct warning
+    };
+    #ifdef __cplusplus
+    }
+    #endif

ggml-blas: refactor BLAS backend #18027

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed

taronaeo wants to merge 24 commits into ggml-org:master from taronaeo:feat/blas_mmid

-Original file line number
+Diff line change
@@ Expand Up / @@ -47,6 +47,7 @@ @@
     /ggml/cmake/                            @ggerganov
     /ggml/include/                          @ggerganov
     /ggml/src/ggml-common.h                 @ggerganov
+    /ggml/src/ggml-blas/                    @taronaeo
     /ggml/src/ggml-cpu/                     @ggerganov
     /ggml/src/ggml-cpu/spacemit/            @alex-spacemit
     /ggml/src/ggml-cuda/fattn*              @JohannesGaessler
@@ Expand Down @@

-Original file line number
+Diff line change
@@ Expand Up / @@ -11,9 +11,10 @@ find_package(BLAS) @@
     if (BLAS_FOUND)
         message(STATUS "BLAS found, Libraries: ${BLAS_LIBRARIES}")
-        ggml_add_backend_library(ggml-blas
-                                 ggml-blas.cpp
-                                )
+        file(GLOB GGML_SOURCES_BLAS "*.c" "*.cpp")
+        file(GLOB GGML_HEADERS_BLAS "*.h" "*.hpp")
+        ggml_add_backend_library(ggml-blas ${GGML_HEADERS_BLAS} ${GGML_SOURCES_BLAS})
         if (${GGML_BLAS_VENDOR} MATCHES "Apple")
             add_compile_definitions(ACCELERATE_NEW_LAPACK)
@@ Expand Down @@

-Original file line number
+Diff line change
@@ -0,0 +1,67 @@
+    #pragma once
+    #include "ggml.h"
+    #include "ggml-impl.h"
+    #include "ggml-backend-impl.h"
+    #include <vector>
+    #include <memory>
+    #include <future>
+    #if defined(GGML_BLAS_USE_ACCELERATE)
+    #   include <Accelerate/Accelerate.h>
+    #elif defined(GGML_BLAS_USE_MKL)
+    #   include <mkl.h>
+    #elif defined(GGML_BLAS_USE_BLIS)
+    #   include <blis.h>
+    #elif defined(GGML_BLAS_USE_NVPL)
+    #   include <nvpl_blas.h>
+    #else
+    #   include <cblas.h>
+    #endif
+    #define GGML_BLAS_NAME    "BLAS"
+    #define GGML_BLAS_VERSION GGML_BACKEND_API_VERSION
+    #ifdef __cplusplus
+    extern "C" {
+    #endif
+    struct ggml_backend_blas_buffer {
+        void * data;  // dequantized data
+        size_t size;  // ggml_nelements * sizeof(float)
+    };
+    struct ggml_backend_blas_buffer_context {
+        void * data;
+        size_t size;
+        std::vector<ggml_backend_blas_buffer *> buffers;
+        ~ggml_backend_blas_buffer_context() {
+            ggml_aligned_free(data, size);
+            for (auto * extra : buffers) {
+                ggml_aligned_free(extra->data, extra->size);
+                delete extra;
+            }
+        }
+    };
+    struct ggml_backend_blas_buffer_type_context {
+        int n_threads;
+    #ifndef GGML_USE_OPENMP
+        std::vector<std::future<void>> tasks;
+    #endif
+    };
+    struct ggml_backend_blas_context {
+        int n_threads;
+    };
+    struct ggml_backend_blas_device_context {
+        char _dummy;  // Prevent empty struct warning
+    };
+    #ifdef __cplusplus
+    }
+    #endif

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ggml-blas: refactor BLAS backend #18027

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!

Uh oh!